Fix fill_value handling during unstack

[peanutfun]

This fixes an issue with the fill_value when performing unstack. The docs on unstack state (emphasis by me):

fill_value (scalar or dict-like, default: nan) – value to be filled. If a dict-like, maps variable names to fill values. If not provided or if the dict-like does not contain all variables, the dtype’s NA value will be used.

If a full reindex is required during unstack, this is performed by the aligner, with the fill value handled here:

xarray/xarray/structure/alignment.py

Line 70 in 6da8c2c

if isinstance(fill_value, dict):

If a full reindex is not required, the emphasized part in the docs is not true. Then, the program flow does not call the aligner but _unstack_once instead here:

xarray/xarray/core/dataset.py

Line 5606 in 6da8c2c

result = result._unstack_once(d, stacked_indexes[d], fill_value, sparse)

_unstack_once, however, does not handle the case where a variable name is not found in the fill_value mapping.

This PR fixes the latter issue.

Please let me know if I should add a test and mention a change in whats-new.rst.

• Closes #xxxx
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst
• New functions/methods are listed in api.rst

[welcome]

Thank you for opening this pull request! It may take us a few days to respond here, so thank you for being patient.
If you have questions, some answers may be found in our contributing guidelines.

[dcherian]

LGTM! Thanks. Can you add a unit test for this please?